A Dynamical System Approach to Robotic Ultrasound Imaging: Towards Intrinsically Stable Robotic Sonography

Wanli Liuchen, Anqing Duan, Zirui Song, Maria Victorova, David Navarro-Alarcon

Abstract: In this research, we introduce a framework that leverages the deep learning capabilities of the Residual Network (ResNet) architecture to improve the efficiency of robotic systems in ultrasound-guided body scanning. It helps accurately spot abnormalities in tissues and understand the detailed aspects of ultrasound images. Subsequently, the framework capitalizes on the strengths of time-invariant dynamical systems (DS) to refine the efficiency of robotic manipulation. Contrasting with conventional approaches necessitating manual probe maneuvering across the patient’s body while maintaining consistent contact force, our methodology permits the robotic arm to fulfill multiple objectives concurrently, such as sustaining uniform contact pressure and achieving exact positioning. By preemptively adjusting the robot’s dynamics to ensure a stable interaction with the patient’s surface, our system addresses minor bodily movements, maintaining constant contact and ensuring a comprehensive and precise scanning process.